Kenya's Senate took its first formal step toward statutory AI regulation on April 2, 2026, when the Artificial Intelligence Bill, 2026 (Senate Bills No. 4 of 2025) received first reading and was referred to the Standing Committee on Information, Communication and Technology. Sponsored by nominated Senator Karen Nyamu and published earlier in February, the bill would establish an Office of the Artificial Intelligence Commissioner and — most consequentially for political speech — criminalise the creation or distribution of deepfakes "intended to deceive, defame, or incite violence," with penalties of up to KES 5 million and two years' imprisonment. The public comment window closed on May 5, 2026.
The case for moving now
The Senate's anxiety is not abstract. Kenya goes to the polls in 2027, and the 2025 information environment already showed the form: AI-generated clips of Interior Cabinet Secretary Kipchumba Murkomen appearing to "resign," manipulated images attributed to Communications Authority leadership during the June 2024 protest period, and bot-amplified disinformation tracked by the Daily Nation and Kenyan researchers. Deepfake harassment of female politicians — including Senator Nyamu herself — is now well documented. A regulator confronting that record can fairly argue that waiting for the courts to retrofit synthetic media into the Computer Misuse and Cybercrimes Act, 2018, is no longer responsive enough. A clear statutory offence aimed at electoral-grade deception, paired with platform-side labelling duties, is the kind of guardrail the EU's AI Act, the UK's draft AI legislation, and Singapore's election-period synthetic media rules have all reached for. The instinct to legislate is not the problem.
What the bill actually does
Beyond the criminal offence, the bill imposes mandatory disclosure on AI-generated content that "resembles existing persons, places, or events" and sets up a risk-tiered registration and audit regime overseen by the new Commissioner. KICTANet's memorandum to the ICT Committee welcomed the bill as "timely" but argued the institutional design — a Commissioner's office rather than a full statutory authority, with no dedicated National AI Fund — is too thin to carry an EU-style risk regime. TechCabal characterised the Commissioner as "a powerful new digital sheriff" empowered to inspect systems, access training data, investigate complaints, and issue enforcement notices, while noting that the qualification floor (a master's degree plus ten-plus years of relevant experience) may exceed Kenya's current AI-policy talent pool.
The carve-out that isn't there
The structural problem is narrower than the headline penalties suggest. As Tech Policy Press has detailed, the bill collapses qualitatively different conduct — deploying a banned mass-surveillance system, skipping a high-risk impact assessment, and posting a satirical AI image of a politician — into a single penalty ceiling. It contains no explicit defence for satire, parody, civic commentary, journalism, or research.
Compare that with the EU AI Act, the model Kenyan drafters explicitly cite. Article 50(4) requires deepfake disclosure but then provides: "Where the content forms part of an evidently artistic, creative, satirical, fictional or analogous work or programme, the transparency obligations set out in this paragraph are limited to disclosure of the existence of such generated or manipulated content in an appropriate manner that does not hamper the display or enjoyment of the work." That is not a loophole for fraud — the deception, defamation, and incitement offences still bite — but it draws a line between propaganda and parody. Kenya's bill draws no such line.
Why this matters in Kenyan context
The chilling-effect concern is not hypothetical. Kenyan authorities have arrested citizens including Billy Mwangi and Rose Njeri over AI-assisted political expression under existing cyber-crime statutes; civic tools that gained traction during the 2024 #RejectFinanceBill protests — "Corrupt Politicians GPT," "Finance Bill GPT" — relied on generative AI to make legislative text legible to ordinary voters. A statute that criminalises distribution of AI-rendered likenesses "intended to … defame" without a carve-out for evidently satirical work hands prosecutors a tool whose scope depends almost entirely on charging discretion. Kenya's Constitution, Article 33, treats political speech as the most protected category of expression; a deepfake offence drafted without that hierarchy in mind sits uneasily with it.
A proportionate fix
The Committee does not need to abandon the bill — and the pro-speech case is not that deceptive synthetic media is fine. It is that the bill's good provisions (a labelling duty on platforms, criminal liability for election-grade fraud, an institutional home for AI oversight) are weakened, not strengthened, by the absence of three small amendments. First, an Article 50(4)-style carve-out for evidently artistic, satirical, journalistic, and research uses. Second, a tiered penalty structure that separates platform compliance failures from individual speech offences — the satirical image and the unaudited surveillance system should not share a sentencing ceiling. Third, an explicit reference to Constitution Article 33 as an interpretive lens for the Commissioner and the courts.
KICTANet has called for upgrading the Commissioner's Office to a full statutory authority — a sensible structural ask. The narrower textual fixes are cheaper still. Done well, Kenya's AI Bill becomes the regional template Senator Nyamu's office has indicated it wants. Done as currently drafted, it risks being remembered as the law that armed the 2027 cycle's prosecutors rather than the law that defended its voters.